Crowdsourced air temperature data for the evaluation of urban microscale simulations: Insights into spatiotemporal patterns from three German cities

Rapid development of microscale urban climate models in recent years requires ongoing model evaluation under different scenarios and conditions. In a previous study, we utilised crowdsourced air temperature data from Netatmo citizen weather stations (CWS) for the evaluation of the PALM model during a hot summer day in the city of Bochum, Germany. The data proved valuable due to their high spatial resolution, even though a temporal pattern in model performance with an underestimation of air temperatures at night was observed. However, this finding was based on a single city and limited episode. In this paper, PALM simulations for three cities with different size and geographical setting in Germany (Dortmund, Cologne, and Berlin) are compared with respective crowdsourcing data to test the potential and robustness of this approach. Each simulation covers a period of three days during an observed heat wave in August 2020. The model evaluation with crowdsourced data reveals a high model performance in all cities. At the same time, temporal and spatial patterns in the differences between the modelled and crowdsourced data can be detected, confirming the underestimation of nighttime air temperature by the model especially in densely built areas, as was found in the precursor study. Nevertheless, the simulations show that the PALM model is capable of compensating a great share of the differences between mesoscale model predictions and observed temperatures. Overall, we find that crowdsourced data is an easily available tool for model evaluation and that the PALM model shows a high accuracy over a range of different cities and geographical settings.